# AI copilot

An AI copilot is an assistant built into a piece of software that works alongside the user. It suggests, drafts, summarizes and answers questions, while the user reviews the output and takes the actions themselves.

## What does an AI copilot do?

An AI copilot speeds up work the user is already doing. The name says it: the copilot helps, the pilot flies.

Typical jobs look like this:

- **Suggest.** Complete the next line of code, propose a reply to a ticket, recommend a formula.
- **Draft.** Write a first version of an email, a document section or a query.
- **Summarize.** Condense a long thread, a meeting or a record history.
- **Answer.** Explain a feature or a number, using the data on screen.

In every case the user stays in control. The copilot produces something, and the user decides whether to keep it, edit it or discard it. That keeps risk low and makes copilots easy to add to almost any product.

## What is the difference between an AI copilot and an AI agent?

An AI copilot assists with the task. An AI agent does the task.

| | AI copilot | AI agent |
|---|---|---|
| Who acts | The user | The agent, within limits |
| Output | Suggestions and drafts | Completed actions, then a summary |
| Scope | The current screen or document | A goal that may span many steps |
| Needs | Context about what you're doing | Tools, permissions and a way to check results |
| Best when | The user wants to stay hands-on | The user wants the outcome |

Take a finance manager who wants every expense report from last month that is missing a receipt sent back to its owner with a note. A copilot can explain where the approval queue is, or draft the note. The manager still opens each report and sends it back. An agent finds the reports, returns each one through the product's own action and reports back with the list.

Many products start with a copilot and move toward agents as they expose more of their actions to the model and gain confidence in permissions and controls.

## When should a product use a copilot or an agent?

Choose a copilot when the work is creative or judgment-heavy (writing, design, analysis) and the user wants to shape every line. Choose an agent when users come for a result and the path to it is clicks, settings and repeated steps they would happily skip.

The second case is where Foldspace works: users ask for the result, and the Product Agent takes the clicks off their hands, doing the steps in the product [through the product's backend APIs](https://docs.foldspace.ai/start/faq/) with the signed-in user's permissions.

## Learn more

- [Product Agent](/platform/product-agent)